Revealing the importance of groundwater for potable private water supplies in Wales

At least 77,000 people across Wales rely on private water supplies for their drinking water, with 94 % of these supplies dependent on groundwater. Potable private water supplies were mapped to Lower Super Output Area Level, creating the first map of its kind for Wales. Some rural areas report nearly 43 % of properties using private water supplies as their principal source of water. Simplifying the complex geology of Wales into ‘hydrostratigraphic units’ shows that 97 % of private water supplies are sourced from secondary aquifers which have low productivity and storage. Ordovician and Silurian bedrock aquifers and associated Quaternary deposits support nearly 75 % of all private water supplies. The total volume of groundwater abstracted by licensed and unlicensed potable private water supplies across Wales is estimated at 24.6 ML /day. In times of drought many of Wales’ low storage aquifers can experience insufficiency events. During 2018, reports of 132 dry supplies were collated, however we suspect many cases were not reported. In a changing climate with more extreme weather events, and as working from home becomes more common, private water supply users in low-storage and low-permeability aquifers may find themselves at increasing risk of insufficiency events. AC CE PT ED M AN US CR IPT Running header: private water supplies in Wales by guest on March 14, 2022 http://qjegh.lyellcollection.org/ Downloaded from


Introduction
Private water supplies (pws) are used for potable drinking water, commercial, industrial, agriculture and leisure purposes, often in areas where connection to the mains public water supply is not possible.
Globally these abstractions provide water to billions of people, and may increasingly be at risk of depletion due to pressures from climate change and abstraction demand (Jasechko and Perrone, 2021).
Whilst perhaps more recognised as water supply sources in low-income nation settings, there are considerable numbers of users in high-income nation settings as well. For example, in Europe, it was estimated that 20 million people (7% of population) use small pws, serving fewer than 50 people, in 27 nations which responded to a survey by the World Health Organization (Rickert et al., 2016).
There are 74,251 registered pws in Great Britain (DWI, 2019a;2019b;DWQR, 2019) of which 14,846 potable supplies are registered in Wales. In Great Britain it is estimated that at least 1.9% of the population or about 1.27 million people use water from pws (DWI, 2019a;2019b;DWQR, 2019).
In Wales over 77,000 people are known to live or work in properties that depend upon pws (DWI, 2017). Analysis of properties that do not fall within the public water mains supply area suggest that there could be as many as 41,760 pws in Wales covering potable, agriculture, tourism and commercial uses (ARUP, 2021). In Britain, private water supplies can be sourced from surface waters, including rivers or lakes, rainfall collectors and groundwater sources, including boreholes, wells, adits and springs.
The majority of potable pws are unlicensed and abstract relatively small volumes of water, however supplies that abstract more than 20m 3 /day are licensed by Natural Resources Wales. Private water supplies to domestic properties are known to be vulnerable to water quality problems (e.g. Ander et al 2016;Crabbe et al 2017;DWI, 2017). However water supply vulnerabilities, and risks from future changes arising from changing climate (CCRA, 2017) are often poorly understood for these types of small supplies in Great Britain. Research in the United States (Murti et al 2016) illustrated that during drought events pws users had reduced their indoor and outdoor consumption.

A C C E P T E D M A N U S C R I P T
Drought has been a recurring feature of the UK climate (Environment Agency 2006) however pws users may face more frequent challenges from a changing climate, with extreme weather events (CCRA, 2017;Murphey et al. 2018) often resulting in insufficiency of supply during droughts, or contamination of supplies during flood events. The effects of drought on water users is rarely considered in temperate climates (Bryan et al., 2020). Predictions of a higher likelihood of drought (Burke et al. 2010) and extreme weather events (CCRA, 2017;Murphey et al. 2018;Kendon et al. 2013) underline the urgent need for a better understanding of the location of pws and the resilience of the aquifers from which they abstract water in order to better protect the people and businesses who depend on them. Spatially predictive models (Mansour et al. 2018) also suggest future variability and change of groundwater recharge, with some areas seeing a reduction and others an increase in effective recharge to groundwater. Effects of insufficiency can be wide ranging and can include; impacts on water supply, water quality, and agriculture (e.g. Cook et al 2017), physical and mental well-being (Welsh Government, 2015) and financial and operational challenges for businesses and home owners. It is possible that insufficiency events will have a greater impact on the elderly, less mobile or otherwise more vulnerable in society. During the COVID-19 pandemic, which saw several lockdowns and many people working from home from mid-March 2020, water use in some areas was noted to have increased by 25% as a combined effect of warm temperatures and people being at home more often (WaterUK, 2020). If the move to working from home, or relocating to more rural areas, becomes more common it is possible that water demand from potable pws could increase.
One challenge to better understanding risks to pws is the absence of publicly available maps describing the location and source of pws at a resolution below that of local authority. This lack of mapping can result in difficulties for local authorities, water companies, government bodies, environmental regulators, planners and developers who all require information about the location to protect pws and their users. To compound this issue no official records of pws impacted by resilience issues (drought) in Wales have historically been collated.
This paper aims to improve the understanding of the location and source of potable pws abstracting from groundwater across Wales, to reveal where these are important in order to underpin management of future drought-induced insufficiency events, including the potential effects of longer-term water stress due to climate change. The first Wales-wide map of pws distribution, estimated abstraction volumes and occurrences of insufficiency following the 2018 drought (Turner et al 2021) is presented.
The importance of groundwater for pws in Wales is described and supplies are attributed to "hydrostratigraphic units" which are geological units with similar hydrogeological characteristics or properties (Loveless et al. 2019).

Study area
Wales had an estimated 1.3 million households and a population of 3,125,200 in 21073,125,200 in (ONS, 2017. Wales covers an area of 20,700 km 2 of which 80 % is used for agriculture (Welsh Government, 2015) and is both topographically and geologically varied. The population is concentrated in coastal cities in the south and in larger towns along the coast in the north and west. Inland, a significant population was established in the south Wales valleys during the development of the extractive coal industry. The interior is mountainous, rural and sparsely populated. Annual rainfall can vary from 3000 mm per year in Snowdonia to 1000 mm per year along the coast (MetOffice, 2021).
Private water supplies provide water to about 77,000 people across Wales from 14,846 registered supplies (Drinking Water Inspectorate, 2019a). Environmental Health Officers at 22 local authorities are responsible for recording, risk assessing and testing the water quality of pws in line with the requirements of The Private Water Supply (Wales) Regulations 2017. The majority of pws (81 %) are classified as "single domestic dwellings" and are often the only water supply to rural and remote properties, many of which cannot access the mains water supply network.
The majority of Welsh aquifers are considered to be of low to medium productivity and as such are classified as "secondary A or B aquifers" previously known as "minor and non-aquifers respectively (Jones et al. 2000). Secondary aquifers cover 94.7 % of Wales" surface area and are considered important for pws, however they can have relatively poor aquifer properties e.g. moderate to low productivity and storage capacity. Groundwater in bedrock aquifers, contained mainly within shallow weathered zones, joints and fractures, generally supports small localised supplies. Quaternary sand and gravel aquifers classified as "secondary A" aquifers, are also used for smaller localised public water supplies (e.g. Hiscock & Paci, 2000). Groundwater for public supply in Wales is abstracted mainly from the principal Carboniferous Limestone and Permo-Triassic sandstone aquifers (Allen et al. 1997) however principal aquifers only cover about 5.3 % of Wales" surface area, in the far north east and south of the country.
Hydrogeological information for aquifers that support pws in Wales is limited (Robins, 1999, Robins et al. 2007Robins et al 2000;Robins & Davies, 2016;Hiscock & Paci, 2000, Jones et al. 2000Jones & Farr, 2015;Moreau et al. 2004;Robins & McKenzie, 2005;Shand et al. 2005, Neal et al 1997 and as a result there is a paucity of data for many of these secondary aquifers and their response to drought is poorly understood. A review of historical drought in rural communities in Wales provides evidence that insufficiency of supply in Wales is not a new occurrence, despite it often being considered a "wet" country (Waddington, 2017). However since the 19 th Century resilience to drought stress has improved (Marsh et al. 2007), possibly due to advances in mitigation, planning and infrastructure (EA, 2006).

Private water supply data
Private water supply records are maintained by Local Authorities, and compiled annually into a national dataset by the Drinking Water Inspectorate (DWI). The DWI supplied the pws records for 2015 with their location aggregated to Lower Super Output Areas (LSOA) (DWI, 2017;Ander et al 2020). The LSOA are geographical delineations, fully nested within other administrative geographies, including that of Local Authority (ONS, 2011), suitable for provision of consistent spatially referenced data. The LSOA represent areas that contain between 400 -1,200 households or 1,000 -3,000 people, and there were 1,909 LSOAs in Wales in the Census 2011. This spatial aggregation therefore ensured that the location of individual supplies, or households, remained confidential during this study in accordance with data protection regulations and best-practice. Further data checks established that 473 pws had not had location records updated for legacy reasons and these had been recorded to the local authority office LSOA. These were removed from further analysis to avoid skewing results. It is acknowledged that there are potentially many unreported pws across Wales, particularly those likely to be used for single domestic dwellings, however only the official records were used as no other data was available from which to estimate the number of missing sites. Nonpotable abstractions for agriculture (e.g., irrigation and livestock watering) or small scale industrial processes were also excluded.
The Private Water Supply (Wales) Regulations 2010, in force for the reported data from 2015,categorised supplies into; single domestic dwellings (no regulation number), small-shared supplies (Regulation 10), large supplies (Regulation 9) and re-distribution networks (Regulation 8), and this terminology is used in this paper. Subsequent changes to the Welsh Regulations (The Private Water Supply (Wales) Regulations 2017) mean that later data is recorded as; re-distribution networks (Regulation 8), large supplies (Regulation 9), single domestic dwellings (Regulation 10), and small shared supplies (Regulation 11). "Supplies" is used specifically to refer to the water sources at point of abstraction, not the properties/households estimated to be using the supplies, which will be larger in number.
The sources of pws in Wales (DWI, 2016) were simplified into three categories: groundwater, including springs (SPW), wells (WEL), boreholes (BHW) and adits, and mixed sources (MMS; MXW); surface water (SFW) including streams, ponds and lakes; and, rainwater (RNW) including roof top and other catchment systems. Missing source data is recorded as unknown (UNK).

Delineating Hydrostratigraphic Units
The bedrock and Quaternary geology of Wales is complex, and a simplified classification was required to support a broad Wales-wide analysis of pws sources. The framework used was that of "hydrostratigraphic units" defined as geological formations, or groups of formations, with similar hydrogeological characteristics or properties (e.g. porosity or hydraulic conductivity) relating to groundwater flow, and potentially other hydrogeological characteristics, such as groundwater chemistry (Loveless et al. 2019).
The hydrostratigraphic units aquifer map was derived from the British Geological Survey"s 1:625,000 scale bedrock geology dataset (Loveless et al., 2019). Ten hydrostratigraphic units were created: Precambrian; Cambrian; Ordovician and Silurian; Devonian; Carboniferous Limestone; Carboniferous Coal Measures; Permo-Triassic sandstones; Triassic Mercia Mudstone; Jurassic (Lias); and, Eocene periods. For the purposes of aquifer management, the Carboniferous Limestone and Permo-Triassic sandstone are classified as principal Aquifers (whilst the remaining geologies are classified as secondary A or B Aquifers (Jones et al., 2000).

Quaternary
Quaternary aquifers were mapped using the British Geological Survey"s 1:625,000 scale superficial deposits productivity map developed by Coxon et al, (2020). Hydrogeological classes for each formation type were derived based upon information on their relative productivity, primary hydrogeological data and dominant lithology. Coxon et al (2020) divide the Quaternary deposits into five classes, however for this analysis the classes were simplified into two groups; "moderately productive Quaternary deposits" (including glacio-fluvial and alluvial sand and gravel deposits) and "all other" Quaternary deposits" (including peat, diamicton (till), blown sands and mudflat deposits) which are considered to be less productive water sources.

Estimating abstraction volumes
The volume of water abstraction for the majority of unlicensed pws is not recorded, so a method of estimating this was developed. Private water supply numbers and types (DWI, 2017) were assigned estimated abstraction volumes. Single domestic dwellings, were estimated assuming a daily usage of 150 L/person (Environment Agency, 2009) and an average occupancy of 2.3 people per property (Office for National Statistics Average household Size Wales; 1991 to 2019). Large pws (Regulation 9) are defined partly on the basis of a minimum abstraction of 10 m 3 /d. The number of households using each small shared supply is not recorded, so a value of 5m 3 /d (half of the maximum permissible small shared volume) was attributed to each supply. The data for re-distribution networks (Regulation 8) were not used as these were assumed to refer to the onward supply of public water supplies. For supplies greater than 20m 3 /d an abstraction licence is required and abstraction data was supplied by Natural Resources Wales for 68 licenced private water supplies, comprising 24 surface water abstractions and 46 groundwater abstractions. The daily authorised abstraction quantity for each supply was used to calculate a total for licensed potable pws in Wales which was added to the estimated volume of unlicensed pws described above.

Spatial data integration and analysis
ArcGIS (v10.3, ESRI) was used for geographical analysis and map production. Spatial joins were used to link the LSOA areas and the hydrostratigraphic units, with the percentage of each hydrostratigraphical unit in each LSOA calculated. For the majority of LSOA only one hydrostratigraphic unit was present, however where the LSOA polygon intersected more than one hydrostratigraphic unit polygon, the pws were allocated pro rata to the relative areas of the aquifers, because location information within the LSOA was not available.

Reports of insufficiency during 2018
Failed (dry) pws were voluntarily reported by householders to each local authority and then collated by Welsh Government. There was no methodical survey to assess the type or location of supplies impacted by drought. Only potable pws were reported, agricultural supplies were not included. To protect users" data they were aggregated at a local authority scale. A total of 132 cases of insufficiency in potable pws were collated by Welsh Government during 2018, with the first recorded during the week commencing (wc) 16 th July and the last in the wc 3 rd September. Although this study focuses on Wales it is worth noting that similar issues were experienced elsewhere in the UK, for example in Scotland (Rivington et al 2020), and in the Republic of Ireland (Mooney et al., 2021).

Results
Of the 14,846 registered pws in Wales, the 94% that were sourced from groundwater (boreholes, wells, springs and adits) were used, equating to 13,657 supplies.

Spatial Mapping of potable Private Water Supplies
The first national maps showing the spatial distribution of potable pws disaggregated within local authority areas are presented ( Fig. 1; 2). These maps show considerable variation in number of records and estimated proportion of households, even within the authority areas with the largest number of supplies. They also suggest boundary effects between the different authority areas (e.g. boundary of Gwynedd and Powys) which are likely due to differences in reporting. The smaller, urban local authority areas show a low density of pws, as most properties are supplied by mains water, especially in south Wales and along the north Wales coast. Higher numbers and percentages of pws are recorded in mid, west and north Wales where populations are more rural and dispersed. Of these, Powys covers the largest area in Wales, has the most recorded pws, and also contains the highest number of pws in a single LSOA with 415 pws (Fig. 1) which equates to 43% of the known properties in the area (Fig. 2).

Allocation of supplies to Hydrostratographic Units
Hydrostratigraphic units for bedrock and Quaternary aquifers (Fig. 3.) were derived from 1:625,000 British Geological Survey data. The pws database does not include borehole lithological logs (recorded at the time of drilling), or if they are well or spring sources, thus it was not possible to identify if the source of groundwater was from the Quaternary or bedrock aquifers or a mixture of both. Unique bedrock hydrostratigraphic units were delineated for 67 % of the LSOA. Greater discrepancy was found in other areas, especially for coastal LSOA which are mapped down to the mean low tide level, and thus include estuaries, coastal blown sands and beaches that are not expected to form potable pws aquifers to any great extent.
The hydrostratigraphic units approach also allows us to make a first estimate of the importance of groundwater abstraction (via springs, wells or boreholes) from formations which have previously considered to hold inconsequential quantities of water. Table 1 shows the distribution of groundwater sourced pws attributed to hydrostratigraphic units. This found that 97 % of pws are estimated to be sourced from areas that are traditionally considered to be poorly productive "secondary" bedrock Wales, underlying the cities of Cardiff and Newport and is classified as a secondary aquifer. Although traditionally regarded as impermeable and at best a poor aquifer, groundwater can be abstracted from occasional sandstones and siltstones that can provide localised water supplies (Jones et al 2000). The marginal facies, found at the base of this group, can also yield useable quantities of groundwater.

Abstraction volume of potable Private Water Supplies in Wales
The abstraction volumes for the majority of the unlicensed pws (Table 2) are at best estimates, as unlike the licenced abstractions monitoring is not required. When broken down into percentages based on the water source the unlicensed private water supply data show that groundwater accounts for 94% of recorded supplies, surface water for 6% and rainwater for <0.5%. Interestingly this is almost the reverse of the public water supply, which is provided by surface water (94 %), with just a small groundwater (6 %) contribution.
Unlicensed potable pws abstractions supply 23.7 ML/d (groundwater), 1.6 ML/d (surface water) and 0.005 ML/d (rainwater) whilst licenced potable pws abstractions supply 0.9 ML/d (groundwater), 1.2 ML/d (surface water) and 0.06 ML/d (rainwater). Combing the licenced and unlicensed data the total volume of potable pws is estimated at 27.5 ML/d, comprising; 24.6 ML/d from groundwater, 2.8 ML/d from surface water and 0.1 ML/d from rainfall (Table 2).
Groundwater abstraction for public water supply is calculated from licensed volumes at 25.7 ML/d, broadly similar to the estimated volume of groundwater abstraction for pws. Combining groundwater abstraction by both potable private and public water supplies it is estimated that Welsh aquifers contribute 50.6 ML/d to Wales" potable water supply.  (Fig. 4). This is likely to be a significant under-reporting of the true number of pws that suffered supply issues in 2018, however there is only anecdotal evidence for this and as such the true number of supplies impacted cannot be quantified.

A C C E P T E D M A N U S C R I P T
The highest number of reports came from Gwynedd (n = 31), Powys (n = 21), Ceredigion (n = 25) and Carmarthenshire (n = 22) which are supported by secondary aquifers. Despite a large number of supplies there were no recorded reports of insufficiency in Anglesey or Denbighshire. It is not possible to draw significant conclusions from the reporting of failed supplies during 2018, and thus the true number of pws at risk, or the most vulnerable aquifers are poorly understood.

The spatial distribution of private water supplies
Presentation of the disaggregated location of pws shows that there is considerable variation in the location of supplies within the local authority areas, and this provides considerably more information at the national-scale than the previously available local authority summaries. In particular these data help to better understand where there may be communities with large numbers of households at risk from water supply challenges, including insufficiency, and support national contingency planning by providing a more nuanced perspective on where particular clusters of households and/or businesses Spatial controls are likely to include distance to mains networks in the lowest population density areas, as seen elsewhere (Johnson & Belitz, 2017). Customers are responsible for all costs associated with connection and maintenance of pipework from the mains pipe, which tends to lead to a decrease in connections as distance increases, and this is why some of the most rural areas of Wales are effectively without mains supply. This new systematic national presentation of the location of pws facilitates communication of information related to aquifer properties, which inevitably cross administrative boundaries, and provide an overview of where communities, residences and businesses vulnerable to insufficiency of water supply are located.
The known absence of some pws sources in the official records, especially for the smallest supplies, is for legacy and resourcing reasons. This means that the disaggregated mapping and derived estimates Using the aggregated approach and reporting to a LSOA it was not possible to distinguish whether individual pws source groundwater from bedrock or Quaternary units, or both, with a high level of confidence. Future analysis could be improved by using the precise location of the pws, although this will still require some assumptions to identify the source aquifer.

The hydrostratigraphic units approach
The hydrostratigraphic units approach provides a framework for communication of the geological context of groundwater-dependant pws within and beyond the geological sciences community. In particular this helps to widen the understanding that geologically-related risks (quantity or quality) cross administrative boundaries, and are designed to build up the opportunity for shared learning between different organisations charged with protection of health or water resources where not all participants have a background in earth-environmental sciences. This hierarchical system also allows new information on specific formations to be readily integrated for dissemination.

Water usage estimates
Water use for the majority of unlicensed potable pws (single domestic dwellings), was calculated assuming an estimated volume per person [150 L/day], for users of public (mains) supplies in the absence of pws specific data. This is a significant data gap, as it is not clear if users of pws use more water, perhaps because they view it as "free" and "unlimited", or whether they are wary of shortages and use less than typical mains users. It is acknowledged that this may not be an accurate reflection of private supply water use, especially in domestic settings. Robins, et al (2000) suggested that pws users in the Teifi Valley used on average 600 L per day, however it was not stated if this was for potable or combined uses including agriculture. Assuming 2.3 people per household this suggests average water usage of 260 L per person per day, however it was not possible to state how much of this was used for potable purposes or how representative this was for pws across Wales.
It is acknowledged that outside of potable water there are many others uses for private water supplies, including; agriculture, leisure or drinking water for livestock. A recent estimate (ARUP, 2021) suggested there could be a total of 41,760 private water supplies with a demand of 104 ML/d for all private water uses in Wales, however similar data gaps on actual volumes of water used were also reported as limiting factors in this analysis.
It will be important to acquire improved water usage and availability data in order to better understand the resilience of the aquifers and users to low recharge or groundwater flow conditions. Given these structural uncertainties of missing data, careful thought should be given to how to control for these in any formal uncertainty calculations in the estimations of potable users, as these may be more important than the currently quantifiable parameters. Incorporation of commercial pws, such as agricultural uses, would be needed in any catchment-or aquifer-specific abstraction water balance calculations or modelling.

Current and future resilience challenges
The total numbers of reported dry pws are low compared to the number of total recorded supplies.
For example in Powys where there are 6,011 registered supplies, only 21 supplies were reported to the local authority as drying up. Whilst this could be taken as suggesting the effects of the prolonged dry weather were not recognised in many aquifers, anecdotal evidence suggests there is likely to be a significant under-reporting of drought impacted pws in Wales. Reporting in national and local media also highlighted the difficulties this creates for people whose water has run dry (BBC News Online, 24 th July 2018).
Under-reporting could be due to users who are "used to" experiencing insufficiency of supply and have appropriate contingency plans in place, e.g. increased storage and therefore may not report a dry water supply to the authorities. In Ireland, 32% of pws users in a survey reported experiencing the 2018 drought there. In the same survey, 27% of respondents reported using alternative water sources during extreme (drought or flooding) weather events (Mooney et al., 2021).
In a changing climate with more extreme weather events, pws users in low storage and permeability aquifers may find themselves at increasing risk of insufficiency. Effective recharge to groundwater may increase, or decrease, or become more concentrated in shorter periods in the future (Mansour et al., 2018) which would impact water resource within many of the shallow fractured aquifer in Wales.
This may not be recognised by pws users themselves -in Ireland 54% of study participants recognised climate change impacts as a concern for their pws groundwater quality (Mooney et al., 2021).
The challenges to water users may be in part due to geography, for example distance from an alternative supply of water, or due to difficulties or expense of a mains connection, but also due to the physical properties of the hydrostratigraphic units, many of which are unable to store large volumes of water. On site storage tanks and other physical measures can help build resilience for pws users, however even these systems can be limited especially after prolonged dry weather events.
Dependence upon secondary aquifers, of which very little is known in Wales, compounds the risk of predicting when insufficiency events will occur and highlights the need for strategic monitoring and reporting of the status of secondary aquifer systems across Wales. The same knowledge gaps exist elsewhere in the UK where there are large numbers of private water supplies: although local water quality studies may provide specific insights (e.g. Ander et al. 2016), the "desk based" approach used here can help provide a framework to place localised studies into (via the hydrostratigraphic unit classification), and provide a basis for galvanising future data collection priorities in these dispersed rural settings.
There are multiple challenges that could compound resilience issues for pws users in Wales.
Changing climate with drier summers and more intense rainfall could result in more frequent insufficiency events (CCRA, 2017), whilst floods could also pollute poorly protected pws. Changing rural economies and farming practices could see water demand increase. Recently the COVID-19 pandemic has seen a move to people working from home, using an estimated 25 % more water (Water UK 2020). As remote technology improves it is unknown if this will result in more people moving to, and working from home in, rural areas. If it does the demand on pws could increase.
The most numerous supplies are single domestic dwellings and metering of a small subset of these supplies could be used to understand water usage and behaviour before, during and after insufficiency events. Voluntary reporting of insufficiency events is unlikely to provide sufficient detail from which to make decisions, and it is recommended that following future drought events, that pws users are questioned as to the type and resilience of their supplies.

Quantifying drought response in Welsh aquifers
Traditionally groundwater resources in Wales have been considered to be poor and of little use for water supply. As a consequence there has been limited work on the hydrogeology of many Welsh aquifers. However, it is exactly due to their poor aquifer properties that they are more vulnerable to drought. Vulnerability to drought in secondary bedrock aquifers, which cover 94.7% of Wales" surface area, may be greater than in principal aquifers (Permo-Triassic sandstone and Carboniferous Limestone) due to a lesser ability to store water. Despite this, groundwater monitoring and research in Wales has focused on the more productive principal aquifers. Of the few studies conducted elsewhere, these would appear to corroborate that drought, and an associated lack of supply, is less considered than aspects of water quality in management strategies of pws users (Murti et al., 2012;Mooney et al., 2021). Drought is perhaps also less considered in water supply terms in a "wet" nation such as Wales (Bryan et al., 2020;Waddington, 2017), which may have contributed to the lower prioritisation of research in this context. At a household level, demographic factors may also influence user actions, as seen in pws flooding and water quality research (e.g. McDowell et al., 2020), and should be considered in future studies alongside hydrogeological and infrastructure assessments. Data from across Europe shows the importance of groundwater for many small and larger pws users (Rickert et al., 2016), and for which climate change may similarly need to be considered in relation to users" security of supply.
To address this knowledge gap pumping test data from boreholes, in secondary aquifers, including Ordovician and Silurian bedrock aquifers, should be collated and analysed, to provide a better understanding of their physical properties and response to prolonged periods without rainfall at any time of year. Hydrographs from existing NRW monitoring boreholes (Jones & Farr, 2015) should be analysed to provide information on their response to drought events, possibly using the Standardised Drought Indices (Bloomfield & Marchant, 2013). Combining this new information an early warning system for drought in Welsh aquifers could be developed, with outputs from the most reliable and informative sites included in the monthly "Hydrological Outlook" (http://www.hydoutuk.net/), an approach which could be replicated in any location with comparable water monitoring infrastructure.
However, given the separate management of groundwater pws in Wales, accessible communication of this information to individual users will be critical for it to be effective in ensuring preparedness for scare water supply.

Conclusions
Groundwater is often considered of less importance in Wales than in some other parts of the UK because it is not the dominant source of public water supplies. However this study has shown that groundwater, especially from secondary aquifers, is vital for supporting pws across Wales. There are also few studies in similar geographic and socio-economic settings elsewhere, indicating a wider knowledge gap.
In Wales about 94 % of unlicensed potable private water supplies are sourced from groundwater (springs, wells and boreholes), 6 % from surface water (rivers, streams and lakes) and <0.5 % from rainfall. This is almost the opposite of the public water supply, which is dominated by surface water supplies (95 %) with a smaller contribution from groundwater (5 %).  This study presents the first Wales wide map of insufficiency of private water supplies following the 2018 drought. In total 131 reported failures were collated and it is possible that there is significant under reporting of insufficiency events for private water supplies in Wales.
Many of the reported cases were in areas that are underlain by secondary aquifers which may be vulnerable to prolonged dry weather events.
 To address the challenges of future drought and to improve resilience, monitoring, analysis and modelling of the response of groundwater systems and the use of private water supplies is required.
 There are considerable environmental and societal data gaps relating to small-scale domestic private water supplies in particular: these need to be better understood if the vulnerability of communities reliant on these sources is to be proactively protected under changing climate scenarios.  Private water supply data from DWI, 2017. Volume estimates produced as described in the methodology with pws broken down into types of supplies with all springs are classed as groundwater abstractions. Licenced public water supply data from Natural Resources Wales also Welsh Water -Dŵr Cymru (2019) and Hafren Dyfrdwy (2019) Water Resources Management Plans. 381 ML/d of water exported water from the Elan Valley to England has been deducted from the total volume of public water supply.   Private water supply data from DWI, 2017. Volume estimates produced as described in the methodology with pws broken down into types of supplies with all springs are classed as groundwater abstractions. Licenced public water supply data from Natural Resources Wales also Welsh Water -Dŵr Cymru (2019) and Hafren Dyfrdwy (2019) Water Resources Management Plans. 381 ML/d of water exported water from the Elan Valley to England has been deducted from the total volume of public water supply.

A C C E P T E D M A
N U S C R I P T